Developers

Private models behind the API you already know.

Use the official OpenAI SDKs against private open-weight models. One base URL, one API key, one model name — and your existing request code keeps working for supported patterns.

Read the Quickstart API reference

Integration at a glance

OpenAI SDK

Base URL: https://api.lirux.ai/v1
API key: LIRUX_API_KEY=sk-…
Model: qwen3-30b-a3b

Everything else — messages, streaming, response parsing — stays as it is.

First request

Three lines of configuration.

Point the client at your endpoint, read the key from the environment, choose a model. That is the integration.

Works with the official openai packages for Python and Node.js
Chat, embeddings and transcription through the standard SDK methods
Server-sent event streaming
Errors in the OpenAI format, so existing handling keeps working

from openai import OpenAI
import os

client = OpenAI(
    base_url="https://api.lirux.ai/v1",
    api_key=os.environ["LIRUX_API_KEY"],
)

response = client.chat.completions.create(
    model="qwen3-30b-a3b",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Summarise our Q3 support tickets."},
    ],
)

print(response.choices[0].message.content)

Request path

Your code calls an API. We run everything behind it.

Runtime, quantization, gateway, authentication, rate limits and monitoring are managed for you — on shared capacity or on your own dedicated node.

Your application
OpenAI-compatible API
Private AI infrastructure
Open-weight model
Response

Endpoints

Chat, embeddings, speech and reranking.

The endpoints a production AI feature needs — from RAG retrieval to call transcription.

Full API reference

Chat completions

POST/v1/chat/completions

Generate responses from chat, reasoning and vision models. Streaming, JSON mode and tool calling on supported models.

Example model: qwen3-30b-a3b

Embeddings

POST/v1/embeddings

Turn text into dense vectors for semantic search, clustering, deduplication and RAG retrieval.

Example model: bge-m3

Audio transcriptions

POST/v1/audio/transcriptions

Whisper-compatible speech-to-text for calls, meetings and voice input. Multipart upload.

Example model: whisper-large-v3-turbo

Extension

Rerank

POST/v1/rerank

Score candidate passages against a query to reorder retrieval results. Cohere/Jina-style request shape.

Example model: bge-reranker-v2-m3

List models

GET/v1/models

List the model IDs your API key can call, in the OpenAI list format.

Looking for something else — Assistants, Batch, image generation?

See the compatibility matrix

SDK compatibility

No proprietary SDK.

OpenAI-compatible endpoints for supported API patterns. If a tool lets you set an OpenAI base URL, it can usually talk to your endpoint.

OpenAI Python SDK
openai
Set base_url and api_key on the client.
OpenAI Node.js SDK
openai
Set baseURL and apiKey on the client.
Any HTTP client
curl · requests · fetch
Plain JSON over HTTPS with a Bearer token.
Frameworks with a custom base URL
e.g. LangChain, LlamaIndex, Vercel AI SDK
Generally work through their OpenAI-compatible provider options. Check the features you use against the compatibility matrix.

Documentation

Everything you need to integrate.

QuickstartFirst request in Python, Node.js or cURL.API referenceParameters, examples and responses for every endpoint.AuthenticationBearer tokens, rotation, scoping, network controls.ModelsModel IDs, availability, pinning and quantization.ErrorsStatus codes, error codes and retry strategy.OpenAI compatibilitySupported, partial and unsupported features.

Migrate from OpenAI

Move compatible workloads without a rewrite.

Change the base URL, the key and the model name. Then evaluate prompt behaviour, tool calling and context length on the open model before you switch production traffic.

Migration guide

before.py

from openai import OpenAI

client = OpenAI(
    api_key=os.environ["OPENAI_API_KEY"],
)

client.chat.completions.create(
    model="gpt-4o-mini",
    messages=messages,
)

after.py

from openai import OpenAI

client = OpenAI(
    base_url="https://api.lirux.ai/v1",
    api_key=os.environ["LIRUX_API_KEY"],
)

client.chat.completions.create(
    model="qwen3-30b-a3b",
    messages=messages,
)

Get an endpoint for your workload.

Tell us which models and endpoints you need. We deploy, configure and hand over a working API key.

Deploy AI Apply for a Pilot